"#6: Engineering Considerations That Product Managers Should Watch out For" & "#4: Roles, Skills and Org Structure For Machine Learning Product Teams"

"6: Engineering Considerations That Product Managers Should Watch out For" & "4: Roles, Skills and Org Structure For Machine Learning Product Teams"

In the world of machine learning product development, there are several key considerations that product managers need to be aware of. These considerations revolve around engineering requirements, data and model dependencies, data collection methods, and the roles and structure of machine learning product teams.

Real-time requirements play a crucial role in machine learning product development. Product managers must determine whether the results of their algorithms can be calculated in advance or if they need to be calculated in real time. This decision affects the design of pipelines and the choice of storage methods. Real-time requirements can further complicate the data collection process, as it raises questions about whether the data should be collected in batches or streamed continuously, and whether it should be pushed or pulled.

Data and model dependencies are another important consideration for product managers. When data is added or modified, it's necessary to determine which models need to be re-run or even re-trained. Additionally, product managers must establish acceptable service level agreements (SLAs) for how quickly these updates should happen. The rate at which data is collected and accumulated also impacts the design of pipelines and the choice of storage methods.

When it comes to the roles, skills, and organizational structure of machine learning product teams, there are several options to consider. Option 1 is to have data science report to engineering. This alignment between disciplines eliminates the need for a clear delineation between data science and engineering skills. Engineers work closely with data scientists to ensure that the models scale and that the quality of results in production meets the requirements.

Option 2 is to have data science report to product. This approach ensures that data science projects are driven by product needs. It creates full alignment on goals and deliverables between data science and product teams. By reporting to product, data science becomes an integral part of the product development process.

Option 3 is to have data science separate from product and engineering. This structure provides visibility to the data science team and makes it more accessible to the entire organization. However, it may result in less alignment between teams, as there is no single decision maker at the top. Joint reporting, where data science reports to both engineering and product, can also be considered to ensure better alignment between teams.

In conclusion, product managers in machine learning product development must be aware of various engineering considerations, data and model dependencies, data collection methods, and the roles and structure of machine learning product teams. To navigate these considerations effectively, here are three actionable pieces of advice:

• 1. Clearly define real-time requirements: Determine whether the results of your algorithms can be calculated in advance or if real-time computation is necessary. This will impact the design of your pipelines and the choice of storage methods.

• 2. Establish clear data and model dependencies: Identify which models need to be re-run or re-trained when data is added or modified. Set acceptable SLAs for how quickly these updates should happen to ensure timely and accurate results.

• 3. Consider different reporting structures: Evaluate the benefits and trade-offs of having data science report to engineering, product, or both. Choose a structure that aligns with your product needs and goals, and promotes collaboration between teams.

By considering these factors and taking these actions, product managers can navigate the complexities of machine learning product development and drive successful outcomes.